Reproduce “ELEVATE: Agentic Controller Acquisition from Task Failure for Quadrupedal Manipulation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shiqi Liu, Radu Corcodel, Ding Zhao
ELEVATE addresses quadrupedal manipulation failures by using persistent task failures to identify missing physical behaviors, train controllers in simulation, and add them to a reusable library. On A02, reaching a target 1.25 m above the floor near a wall, it achieved 100.0% success across five runs, versus 0.0% for ASPIRE and Code-as-Policies. The comparison includes extra training; physical transfer was task-specific.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.